Make traffic jams a thing of the past – AI traffic light system could significantly reduce congestion

A new artificial intelligence system developed by researchers at Aston University significantly outperforms all other methods.

A new artificial intelligence system reads live camera footage and adapts lights to compensate

In 2014, Americans spent 6.9 billion hours trapped in traffic. During traffic jams, the average commuter used an extra 19 gallons of gasoline. That’s $160 billion in wasted time and fuel every year.

In many major US cities, traffic can waste more than 100 hours a year for the typical driver. In a typical workplace, that’s enough time to take two and a half weeks off. Fortunately, researchers are working to reduce traffic congestion, whether through the development of driverless cars or the use of artificial intelligence in traffic lights.

For example, long queues at traffic lights could be a thing of the past thanks to new artificial intelligence (AI) technology from Aston University researchers. The system, the first of its kind, analyzes live video footage and adjusts lights to compensate, keeping traffic moving and reducing congestion.

The method uses deep reinforcement learning, in which the software recognizes when it’s not performing well and tries a new approach – or continues to improve as it progresses. The system outperformed all other testing approaches, which often depend on manually designed phase transitions. Inadequate timing of traffic lights is a major cause of congestion.

Traffic Light AI System

The new artificial intelligence traffic light system could make traffic jams a distant memory. Credit: University of Aston

The researchers built a state-of-the-art photo-realistic traffic simulator, Traffic 3D, to train their program, teaching it to handle different traffic and weather scenarios. When the system was tested on a real intersection, it later adapted to real road intersections although it was fully trained on simulations. It could therefore be effective in many real-world contexts.

Dr Maria Chli, a computer science lecturer at the University of Aston, explained: “We set this up as a traffic control game. The program gets a “reward” when it drives a car through an intersection. Every time a car has to wait or there is a traffic jam, there is a negative reward. There is actually no contribution from us; we just control the reward system.

Currently, the main form of traffic light automation used at intersections relies on magnetic induction loops; a wire is on the road and records cars passing over it. The program counts this and then reacts to the data. Because the AI ​​created by the University of Aston team ‘sees’ the high traffic volume before the cars have passed through the lights and then makes its decision, it is more responsive and can react faster .

Dr George Vogiatzis, Lecturer in Computing at Aston University, said: “The reason we based this program on learned behaviors is that it can understand situations that it hasn’t. explicitly known before. We tested this with a physical obstacle that causes congestion, rather than traffic light phasing, and the system still worked well. As long as there is a causal link, the computer will eventually figure out what that link is. It is an extremely powerful system.

The program can be configured to visualize any crossroads – real or simulated – and will start learning on its own. The reward system can be manipulated, for example, to encourage the program to let emergency vehicles through quickly. But the program always learns on its own, rather than being programmed with specific instructions.

The researchers hope to start testing their system on real roads this year.

Reference: “Fully autonomous and vision-based traffic light control: from simulation to reality” by Deepeka Garg, Maria Chli and George Vogiatzis, 2022, Proceedings of the 21st International Conference on Autonomous Agents and Multi-Agent Systems.

Comments are closed.